What is behind homicide gender gaps in Mexico? A spatial semiparametric approach

From 1990 to 2016, more than 425,000 homicides have occurred in Mexico. In 2016, the country recorded the world's second largest number of conflict fatalities with more than 24,000 killings. Despite the growing attention for inquiring into specific causes of homicide victimization, research on the matter suffer from three important shortcomings: disregard for introducing a gender perspective, lack of a multilevel approach -use of information both at the victim and community levels- and the exclusive use of traditional linear models -failing to capture nonlinear relationships, as can be expected, in the linkage between age of a person and their likelihood of being victim of crime-. In order to contribute to the analysis of homicides in Mexico, the present study develops a semiparametric approach to investigate the determinants of gender bias in homicide victimization in Mexico. Homicide statistics from 2010 to 2014 and 2010 census data are used to construct a logistic model with sex of the victim as response variable and a set of potential categorical and continuous covariates. The main results suggest that gender differences in victimization can be explained by the mechanism of killing, interaction between age of the victim and the killing mechanism, social deprivation of the municipality of occurrence, share of the population living in female-headed households, share of the population living in indigenous-headed households, random effects and spatial effects.